Organismal, genetic, and transcriptional variation in the deeply sequenced gut microbiomes of identical twins

Abstract

We deeply sampled the organismal, genetic, and transcriptional diversity in fecal samples collected from a monozygotic (MZ) twin pair and compared the results to 1,095 communities from the gut and other body habitats of related and unrelated individuals. Using a new scheme for noise reduction in pyrosequencing data, we estimated the total diversity of species-level bacterial phylotypes in the 1.2-1.5 million bacterial 16S rRNA reads obtained from each deeply sampled cotwin to be approximately 800 (35.9%, 49.1% detected in both). A combined 1.1 million read 16S rRNA dataset representing 281 shallowly sequenced fecal samples from 54 twin pairs and their mothers contained an estimated 4,018 species-level phylotypes, with each sample having a unique species assemblage (53.4 +/- 0.6% and 50.3 +/- 0.5% overlap with the deeply sampled cotwins). Of the 134 phylotypes with a relative abundance of >0.1% in the combined dataset, only 37 appeared in >50% of the samples, with one phylotype in the Lachnospiraceae family present in 99%. Nongut communities had significantly reduced overlap with the deeply sequenced twins' fecal microbiota (18.3 +/- 0.3%, 15.3 +/- 0.3%). The MZ cotwins' fecal DNA was deeply sequenced (3.8-6.3 Gbp/sample) and assembled reads were assigned to 25 genus-level phylogenetic bins. Only 17% of the genes in these bins were shared between the cotwins. Bins exhibited differences in their degree of sequence variation, gene content including the repertoire of carbohydrate active enzymes present within and between twins (e.g., predicted cellulases, dockerins), and transcriptional activities. These results provide an expanded perspective about features that make each of us unique life forms and directions for future characterization of our gut ecosystems.

Fig. 1.

Measurements of bacterial diversity in the human fecal microbiota. (A) Rarefaction curves at 97%ID and 95%ID phylotype cutoffs are shown for the deeply sequenced TS28 and TS29 MZ cotwin (“Deep Twins”) datasets. Sequences were classified as chimeric at the 50% probability cutoff. (B) Comparison of diversity within and between gut microbial communities. Curves at 97%ID phylotype cutoff are shown for 250 fecal samples taken from 146 individuals (“Shallow twins”; 1,000 16S rRNA gene sequences were randomly selected from each sample), 250 samples taken from multiple body habitats (“Whole body”; 1,000 randomly selected sequences per sample), and the two deeply sequenced fecal samples (“TS28-Deep” and “TS29-Deep”). Phylotypes found in multiple fecal samples are labeled “co-occurring.” (C) Plot of proportion of 97%ID phylotypes found in TS28 and TS29 across 277 fecal samples (black circles) and 814 samples taken from multiple body habitats in nine individuals [habitat groups are colored green (fecal), purple (skin), red (external auditory canal; EAC), blue (hair), orange (nostrils), and light blue (oral cavity)]. Four EAC and one skin sample did not contain any shared phylotypes with TS28 and TS29. (D) The proportion of the 250 fecal samples containing each 97%ID phylotype plotted as a function of the relative abundance (%) of each phylotype in the combined dataset. Phylotypes are colored according to phylum: Bacteroidetes (red), Firmicutes (green), and other (black). The expected proportion of samples containing each phylotype, assuming a random distribution across samples, is shown (median ± 95% confidence interval).

Fig. 2.

Diversity of the human fecal microbiome and its metatranscriptome. (A) Distribution of gene clusters across gut microbial genomes and microbiome bins. All protein sequences from 122 gut genomes and the microbiome bins were clustered using cd-hit at 60%ID. (B) Number of sequence variants in each microbiome bin (values normalized by Gbp in bin; all genus-level bins with >100 scaffolds are shown). (C) Rarefaction analysis of the number of genes, gene clusters, expressed genes, and expressed gene clusters in the fecal microbial communities of TS28 and TS29 as a function of sequencing depth. The total number of protein-coding genes in the set of 122 gut genomes and the microbiome bins is 525,329, representing 257,823 gene clusters. (D) Ratio of gene expression to gene abundance (relative abundance of cDNA sequences divided by relative abundance of DNA sequences) mapped to a subset of the bacterial taxa in the fecal microbiome. Taxa with >1,000 mapped cDNA and DNA sequencing reads in both samples are shown.

Fig. 3.

Clustering of fecal microbiome bins and the annotation of differentially expressed genes. (A) UPGMA clustering was performed on the relative abundance of CAZy families across each microbiome bin. Number of genes assigned to each CAZy family was normalized to the total number of genes in each sequence bin (all bins with >30 CAZy family assignments in both samples are shown). Black circles represent clustered nodes after z-score normalization across all bins (inconsistency threshold = 0.75, “cluster” function in Matlab v7.7.0). (B) Percentage of genes with high relative expression (High-Expr) or low relative expression (Low-Expr) assigned to each COG category. Percentages are represented by the area of each circle (black circle labeled 5% provides reference).